|
Home > Archive > Slony1 PostgreSQL Replication > September 2005 > Forcing an existing node to copy data fresh
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Forcing an existing node to copy data fresh
|
|
| Jim Archer 2005-09-07, 8:25 pm |
| Hi All...
I tried posting this question before, but it keeps getting trapped. I
think it's because of my use of the word S*U*B*S*C*R*I*B*E in the message
or in the subject, so here is a new try.
While testing last night, I managed to get a node de-synced due to the wa=
y
I started a slon daemon. I figured that would be a good opportunity to
learn how to make a node start itself from scratch.
>From the docs, it seemed that the S*U*B*S*C*R*I*B*E SET command would do
this if I issued it for a nore that was already inthe cluster. I issued
the command, but nothing happened.
So, I am wondering (1) is this command should cause a working node to
recopy all the data from the provider node and, if not, (2) is there a wa=
y
to make a node do that?
Thanks very much...
| |
| Christopher Browne 2005-09-07, 8:25 pm |
| "Jim Archer" <jim- p8U4U2pV1R2sTnJN9+BG
Xg@public.gmane.org> writes:
> Hi All...
>
> I tried posting this question before, but it keeps getting trapped. I
> think it's because of my use of the word S*U*B*S*C*R*I*B*E in the message
> or in the subject, so here is a new try.
>
> While testing last night, I managed to get a node de-synced due to the way
> I started a slon daemon. I figured that would be a good opportunity to
> learn how to make a node start itself from scratch.
>
> this if I issued it for a nore that was already inthe cluster. I issued
> the command, but nothing happened.
>
> So, I am wondering (1) is this command should cause a working node to
> recopy all the data from the provider node and, if not, (2) is there a way
> to make a node do that?
No, the "expected result" of submitting "that command" to a node
already participating in a subscription is for the subscription
provider to be revised.
Thus, you might start with...
A ---> B ----> C
And revise it to...
A ---> B
\
\
\> C
If you request a 'path' that is already in effect, it'll turn out as a
NOOP.
What you could do to accomplish what I think you want to do is to
submit "*UN*S*U*B****** SET."
--
(reverse (concatenate 'string "ofni.sailifa.ac" "@" "enworbbc"))
<http://dev6.int.libertyrms.com/>
Christopher Browne
(416) 673-4124 (land)
| |
| Jim Archer 2005-09-08, 3:24 am |
| Christopher Browne said:
> No, the "expected result" of submitting "that command" to a node
> already participating in a subscription is for the subscription
> provider to be revised.
>
> Thus, you might start with...
>
> A ---> B ----> C
>
> And revise it to...
>
> A ---> B
> \
> \
> \> C
>
> If you request a 'path' that is already in effect, it'll turn out as a
> NOOP.
>
> What you could do to accomplish what I think you want to do is to
> submit "*UN*S*U*B****** SET."
I just gave this a try. I rebuilt my cluster from scratch and had a 2
machine cluster. I issued:
unsub*scr*ibe set (id =3D @SET_FULL, receiver =3D @NODE_02);
This seemed to work as documented. The database on the slave was
unlocked, so I could modify it, and the slony-added schema remained in
place.
I then issued:
sub*scr*ibe set (id =3D @SET_FULL, provider =3D @ORIGIN, receiver =3D @NO=
DE_02,
forward =3D no );
That didn't seem to do anything. The database remained unlocked and I
didn't see any copy activity in the slon log that services that node.
I then decided to add a third node to the cluster. I issued:
store node (id=3D@NODE_04, comment =3D 'Node 04 Hartford', spoolnode=3Dfa=
lse,
event node =3D @ORIGIN);
This created the schema on that node correctly. I then started a slon to
service this new node and issued:
subscribe set ( id =3D @SET_FULL, provider =3D @ORIGIN, receiver =3D @NOD=
E_04,
forward =3D no);
Again, nothing. No copy activity.
Oh, I created the store paths when I created the cluster originally.
Did I miss something or is there a problem with subscribing sets on an
existing cluster?
| |
| Jim Archer 2005-09-08, 3:24 am |
| Christopher Browne said:
> No, the "expected result" of submitting "that command" to a node
> already participating in a subscription is for the subscription
> provider to be revised.
>
> Thus, you might start with...
>
> A ---> B ----> C
>
> And revise it to...
>
> A ---> B
> \
> \
> \> C
>
> If you request a 'path' that is already in effect, it'll turn out as a
> NOOP.
>
> What you could do to accomplish what I think you want to do is to
> submit "*UN*S*U*B****** SET."
I just gave this a try. I rebuilt my cluster from scratch and had a 2
machine cluster. I issued:
unsub*scr*ibe set (id =3D @SET_FULL, receiver =3D @NODE_02);
This seemed to work as documented. The database on the slave was
unlocked, so I could modify it, and the slony-added schema remained in
place.
I then issued:
sub*scr*ibe set (id =3D @SET_FULL, provider =3D @ORIGIN, receiver =3D @NO=
DE_02,
forward =3D no );
That didn't seem to do anything. The database remained unlocked and I
didn't see any copy activity in the slon log that services that node.
I then decided to add a third node to the cluster. I issued:
store node (id=3D@NODE_04, comment =3D 'Node 04 Hartford', spoolnode=3Dfa=
lse,
event node =3D @ORIGIN);
This created the schema on that node correctly. I then started a slon to
service this new node and issued:
sub*sc*ribe set ( id =3D @SET_FULL, provider =3D @ORIGIN, receiver =3D @N=
ODE_04,
forward =3D no);
Again, nothing. No copy activity.
Oh, I created the store paths when I created the cluster originally.
Did I miss something or is there a problem with sub*scr*ibing sets on an
existing cluster?
| |
|
|
| Jim Archer 2005-09-08, 8:24 pm |
| Tim Goodaire said:
> Often, if you're sending events through and they're not showing up on
> all of the nodes, it's one of the following:
>
> 1. A slon isn't running
> 2. Your paths are wrong
> 3. Your listen network is wrong
> 4. You haven't waited long enough for the event to propogate through to
> all of the nodes
Hm, on Slony 1.1, do I have to manually specify the listen network? I
thought that was handled automagically from the path statements. If not,
that could be my problem...
Thanks...
| |
| Brad Nicholson 2005-09-08, 8:24 pm |
| Jim Archer wrote:
>Tim Goodaire said:
>
>
>
>
>Hm, on Slony 1.1, do I have to manually specify the listen network? I
>thought that was handled automagically from the path statements. If not,
>that could be my problem...
>
>
>
It is handled automatically in 1.1. It's still worth having a look to
see if it's setup properly, though.
--
Brad Nicholson 416-673-4106
Database Administrator, Afilias Canada Corp.
|
|
|
|
|